You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I believe it is simply all the numba compile times. You can try turning on numba compilation caching with environment variables and see if that helps any.
Thanks for confirming. I couldn't find a global toggle in the Numba docs to instruct numba to cache everything. But I added cache=True to every jit decorator call, and that brings down import time from 25s to 8s in a new session.
Hi - importing
fast_hdbscan
seems to be really slow, using the latest PyPi version. It takes ~25s in a new python session.Here is a repro in google colab.
Of course, numba compile times can be really slow sometimes, I just wanted to check if this is expected, or a possible regression due to a dependency?
The text was updated successfully, but these errors were encountered: